NVIDIA has expanded its open-source CUDA-Q quantum computing platform with CUDA-Q Logical, a new orchestration layer focused on fault-tolerant quantum computing. The goal is to let researchers work directly with logical qubits, rather than just physical-qubit toy systems, and to evaluate full-stack quantum architectures much faster.
From physical to logical qubits
In fault-tolerant designs, a single logical qubit is encoded in many physical qubits plus quantum error-correction codes. This overhead is the price for reliable, large-scale computations in fields like materials science, drug discovery and financial modeling.
CUDA-Q Logical provides a unified environment to:
- model quantum algorithms and circuits at the logical-qubit level
- explore different error-correction schemes and code distances
- co-design QPU architectures, control hardware and runtimes
- estimate physical-qubit counts and resource requirements
According to Timothy Costa, NVIDIA’s VP and GM for quantum, the new layer is meant to be an open, customizable platform that can represent all aspects of a fault-tolerant system, regardless of qubit type or architecture, and to accelerate the path to useful quantum–GPU supercomputing.
Early adopters include Fermilab, Infleqtion, IQM Quantum Computers, QCDesign, Quantum Motion and Sandia National Laboratories.
Real-world speedups and QUOPS benchmark
Fermilab used CUDA-Q Logical to compare error-correction approaches, runtimes and physical-qubit requirements. NVIDIA reports that a typical five‑month workflow for developing a fault-tolerant architecture was reduced to about three weeks – roughly a 7× speedup.
Iceberg Quantum applied the platform to Diraq’s silicon spin qubits, finding that around 150,000 physical qubits could support 1,000 logical qubits – about 10× fewer than Diraq’s earlier estimates.
CUDA-Q now also integrates QUOPS, a hardware-independent benchmark from Sandia National Laboratories for tracking progress toward practical fault-tolerant quantum computing. Initial QUOPS results have been published for QPUs from Google, IBM and Quantinuum, and a reference implementation is available through CUDA-Q.
CUDA-Q Logical is available today via the NVIDIA CUDA-Q GitHub repository.
Source: eeNews Europe
